Notion for Teams: A Deep Dive After 9 Months at SoftPortal

TL;DR: We ran Notion Business for 15 people at SoftPortal from August 1, 2025 through April 30, 2026. Built 23 databases, 412 pages, 4 team workspaces. Notion replaced Confluence (we ditched it in October), partly replaced Google Docs (still use Docs for shared editing during live meetings), and tried to replace Linear (we gave up; Notion is not built for engineering issue tracking). The Notion AI add-on at $10 per user per month surprised me by actually paying off in week 6. Search is still slow and the migration trap on October 14 cost us 4 hours of restoration work. Worth $20 per seat per month for any team that values docs, runbooks and lightweight project tracking together.

Jump To

• How We Tested
• What We Built
• Daily Use
• Performance and Cost
• Pros and Cons
• Who This Is For
• Bottom Line

How We Tested

Team size grew from 9 to 15 across the period. Plan: Notion Business at $15 per user per month on annual, $18 monthly. We added Notion AI at $8 per user per month on annual ($10 monthly) on October 4, 2025 after a 14-day trial. Tools tracked: workspace activity logs (Notion's own admin dashboard), Toggl for time spent in tool, weekly 1 to 10 satisfaction poll on Friday afternoons, search query log (we wrote a Chrome extension that logged search-and-then-give-up moments because Notion does not surface this in analytics). Migrated content sources: Confluence (124 pages), Google Docs (about 180 docs), one Asana wiki (yes, we had a duplicate one), and a Trello board's archived cards. Browsers: 9 Chrome, 4 Firefox, 2 Safari. Mobile: 12 used the iOS app daily by month 3; 3 Android users complained about the Android app in week 14 and switched to web on mobile.

What We Built

23 databases. The shape of these is what made or broke our adoption, so it is worth describing. Project Tracker: 1 row per active project, properties for Lead, Status, Quarter, Linked Goal, Risk level, Linked Documents (a relation to the Docs database). Docs: 1 row per important document, properties for Type (Runbook, Spec, Postmortem, OKR, ADR), Owner, Last Reviewed, Status. Decisions Log: every meaningful decision we made as a company; this turned out to be the most-used database after month 3. People: 1 row per teammate, used for ownership relations on every other database. Incidents: postmortems with status, severity, resolution. Customer Pain Log: 1 row per piece of customer feedback, tagged by feature area and severity. Roadmap: linked to the Projects database with quarter and theme. The pattern: build databases for things that have a row identity (a project, a document, a person) and use linked-relations to connect them. Avoid pages that are isolated; they get lost in search and lose context.

The migration trap. On October 14, 2025 I imported 124 Confluence pages using Notion's official Confluence importer. The import looked clean. We started using the new pages immediately. Three days later, on October 17, someone realised that internal links between pages had silently dropped during the import. The Confluence importer maps page-tree relations but does not rewrite inline links that pointed to the old Confluence URLs. About 80 percent of our inline cross-references became dead. The fix took 4 hours: a Python script using the Notion API to find every page, parse for old Confluence URLs in the body, look up the corresponding new Notion page ID by title, and rewrite the URL. Notion's official docs do not mention this. If you migrate from Confluence, expect to write a link-fixer script. If you do not write code, plan to spend a half day manually fixing the most-linked pages.

Daily Use

Notion AI surprised me. I tried it for 14 days expecting it to be a writing assistant for blog drafts. It is not great at that. What it is great at: summarising a long page, extracting action items from a meeting note, or generating a draft postmortem from raw incident logs. Three high-leverage workflows we adopted by week 6. First, weekly project summaries. Each project lead drops their week's notes into a page, runs the AI summarise block, and posts the output to a shared Notion update page. Cuts the time to write a status update from about 25 minutes to 8 minutes. Second, postmortem drafts. Paste the incident timeline from Slack into a page, run an AI prompt with our standard postmortem template, edit the output. Cuts the time from 90 minutes to about 35. Third, customer feedback synthesis. Dump 30 pieces of customer feedback into a page and ask for the top 5 themes. This one is more hit-and-miss but useful as a starting point. AI is the only paid add-on I would keep if we tightened the budget.

Search is the one place Notion has not improved enough. Searching for a phrase that appears in 5 documents returns those 5 documents in an order that is roughly relevance but visibly poor. Our Chrome extension logged 312 search-and-give-up moments across the 9 months, with a clear spike (about 40 in a single week) right after we crossed 300 pages in November 2025. The mitigations we use: explicit Pinned pages at the top of every team workspace, a Quick Links database that anyone can add to, and consistent emoji prefixes in page titles (for example, every runbook starts with the wrench emoji) so type-aheads work. None of these fully fix it. Performance is acceptable but not fast. Pages load in 0.8 to 2.4 seconds depending on database query complexity. A page with a single embedded database view that displays 200 rows takes 3 to 5 seconds and creates noticeable lag when scrolling. Heavy database pages are the worst, especially on Safari. Chrome and Firefox handle them better. Mobile is fine for read, painful for edit.

• Win: AI summarise saves about 17 minutes per weekly status update per project lead
• Win: relation properties between databases create the cross-context we lost when leaving Confluence
• Win: page version history saved us once on December 3 when a deleted block needed restoring
• Gripe: search across 300+ pages is unreliable; mitigations exist but partial
• Gripe: Confluence importer breaks inline links silently; budget time to fix

Performance and Cost

Cost for 15 people over 9 months. Business plan at $15 per seat per month annual, $18 monthly. AI add-on at $8 per seat per month annual, $10 monthly. We started on Business in August at 9 seats and added 6 seats in batches as we hired. Total billed across 9 months: $3,915 for Notion Business plus $1,544 for the AI add-on. Roughly $5,459 in total, or about $40 per user per month all-in. Compare against Confluence Premium at $11.05 per user per month plus Atlassian Intelligence at variable cost (we paid about $17 per user per month all-in before the move) and Coda Pro at $30 per user per month plus AI extras. So Notion sits between Confluence (cheaper) and Coda (more expensive). The honest comparison: Confluence was cheaper but worse at relational data; Coda is closer to Notion but its database performance was slower in our June 2025 trial. Notion is the right midpoint for a 10 to 50 person team.

Pros and Cons

• Pro: relation properties between databases beat every competitor for connected docs
• Pro: AI summarise and postmortem-draft workflows save real hours per week
• Pro: page version history is the safety net for a self-edit culture
• Pro: API is decent for automation and one-off scripts
• Con: search is unreliable past 300 pages and the mitigations are workarounds
• Con: Confluence import silently breaks inline links; allocate a half day to fix
• Con: mobile editing is painful, particularly database views on small screens
• Con: not a substitute for issue trackers; engineering should use Linear or Jira alongside

Who This Is For

Pick Notion if your team is 10 to 50 people and your work is part docs, part lightweight project tracking, part knowledge base. Pick Notion if you want one tool to replace Confluence plus parts of Google Docs and your tracker of choice. Pick Notion if your team writes runbooks, ADRs, decision logs and postmortems and wants those connected by relations. Pick Notion AI if you regularly summarise pages or draft repeating documents. Skip Notion if your primary need is engineering issue tracking; use Linear or Jira and a dedicated tool. Skip Notion if you have heavy live-collaboration needs during meetings; Google Docs is still better for that real-time multi-cursor experience. Skip Notion if your team is under 5 and most work is solo notes; Obsidian or Apple Notes will be faster and free. Skip Notion if you are in a regulated industry that needs audit trails Notion still does not fully provide on Business tier.

Build databases for nouns that have row identity. Pages without a database home get lost. That single rule shaped 9 months of Notion work.

Bottom Line

Nine months in, Notion is the operating system for our company that is not in our code repository. The decision log alone is worth the spend; we cite it in meetings two or three times a week and avoid re-deciding things. The AI add-on pays for itself if you write weekly summaries or postmortems often. Search is the one thing I would fix if I worked at Notion. We mitigate it with tag conventions, pinned pages and the Quick Links database. The Confluence migration trap is real and budget for a half day to fix it if you are coming from there. Will we still use Notion at 50 people? Probably yes, but we will likely move some engineering work to Linear formally rather than half-tracking it inside a Notion database. Got a Notion taxonomy question? Drop me a note. I will share our database schema and the relation patterns that worked for us across 23 databases.